|
Abstract:
|
There is growing interest in meta-analysis of functional magnetic resonance imaging (fMRI) studies. Typical fMRI meta-analyses data consist of activation centers from several studies. Most published methods only produce null-hypothesis inferences and don't provide interpretable, fitted model. To overcome these limitations, we propose a hierarchical marked Bayesian spatial Cox cluster process (HMCCP) model. The posterior intensity function provides information on the most likely locations of population centers (PCs) as well as the inter-study variability of activation centers about the PCs. We model the activation centers as offspring of latent realizations of a parent process for each study. And the study-level parent processes are the offspring of latent realizations (the PCs) of a grandparent process. Also, our model incorporates weights for each study and addresses publication bias.
|
